Fig. 5
From: Exposed soil and mineral map of the Australian continent revealing the land at its barest
Distance between all the National Geochemical Survey Australia soil spectras to the Barest Earth and vegated spectras at the same locations. The cosine distance between the Barest Earth spectra and the NGSA soil sample (a) and between a clear observation exhibiting a high vegetation signal, denoted VEG, and the NGSA soil sample (b). We show this distance as a scatter plot with respect to a local vegetation index chosen as the maximum NDVI in a 75-m2 neighbourhood (3 × 3 pixel) around the location. This shows, as one would expect, that areas with low vegetation have a small distance between the spectra of the Barest Earth, the clear observation and the NGSA soil sample. And vice versa, in areas with high amount of local vegetation we have a higher chance of obtaining a smaller distance between the Barest Earth spectra and the NGSA sample compared to the clear observation. This due to the fact that our Barest Earth model can never obtain a bare soil estimate in areas of permanent vegetation, for example. On the right, we show a map displaying the sample locations coloured based on the the difference between the cosine distance of the Barest Earth spectra to the NGSA spectra and the cosine distance of the VEG spectra compared to the NGSA spectra. We see that the areas showing the largest improvement are in the vegetated areas of Australia located outside the arid area center. We note that the NGSA dataset does not contain samples from some areas of Australia due to land access restrictions.
